Algorithmic Identification of Looted Archaeological Sites from Space

In response to widespread looting of archaeological sites, archaeologists have used satellite imagery to enable the investigation of looting of affected archaeological sites. Such analyses often require time-consuming direct human interpretation of images, with the potential for human-induced error. We introduce a novel automated image processing mechanism applied to the analysis of very high-resolution panchromatic satellite images and demonstrate its ability to identify damage at archaeological sites with high accuracy and low false-positive rates compared to standard image classification methods. This has great potential for large-scale applications whereby countrywide satellite datasets can be batch processed to find looting hotspots. Time is running out for many archaeological sites in the Middle East and elsewhere, and this mechanism fills a needed gap for locating looting damage in a cost and time efficient manner, with potential global applications.